AVX? (Other new features?)

Dubslow · 2011-10-14, 02:37

Sorry if this is redundant somehow, but there's not been a new release since I've been around, and I'm curious how often they are. More specifically, I remember seeing in some milestone thread a while back that George mentioned he got his first AVX FFT working, so I'm wondering when that will be implemented, now that the relevant procs are out? And in this vein, when will we see mfaktc/o integration? (mfakto would make more sense.) Could it be something in the next 6 months to a year, or later or never?

Thanks
Dubslow

cheesehead · 2011-10-15, 20:24

GIMPS is an all-volunteer effort. Many of the volunteers have full-time jobs doing other things.

The work will get done when it gets done. Quit asking for specificity.

Are you willing to do part of the work? Then you can issue your own time estimates for your part. The source code is available for you to peruse. You could outline your qualifications to George Woltman and ask what task you might help perform.

Are you willing to pay someone to get something done sooner than the all-unpaid-volunteer pace would deliver it?

cheesehead · 2011-10-15, 21:12

Quote:

Originally Posted by Dubslow

Sorry if this is redundant somehow, but there's not been a new release since I've been around, and I'm curious how often they are.

How often? Whenever they occur.

In the early days, when fixing major bugs and adding new features were more frequent than they are now, releases were more closely-spaced than they are now.

Dubslow · 2011-10-15, 23:02

Quote:

Originally Posted by cheesehead

In the early days, when fixing major bugs and adding new features were more frequent than they are now, releases were more closely-spaced than they are now.

That's as much as I was asking for.
As for the new features, just a ballpark estimate of how hard it would be. Like, easy, easy-with-problems, or hard. I'd help, except I'm not a programmer, and therefore don't know how difficult these things are. (I think I will be taking a Java/OOP course next semester.)

Christenson · 2011-10-16, 02:04

Java is going to be a tough thing to learn if it's your first programming language.

Doing the communcations automagically in mfaktc is supposedly my project. It's been on the burners for six months; history suggests it will be another six months before get it finished...because of that full-time job I do, which had me working this morning in quite a bit of pain rather than sleeping in properly. The outline of the project so far has been to upgrade the worktodo parser, then I will start figuring out how to link in P95 routines from mfaktc so as to re-use as much of P95 as possible.

One suggestion is you take a look at mfaktc and understand how it produces the pseudoprimes that then get passed to the GPU. We are going to want to move that task onto the GPU. Another is to help the gentleman doing the public display figure out what he wants to in terms of diagnostics out from either P95 or mfaktc. I think he's going to want to do a creative display of the FFT data being operated on. Once you have started looking at mfaktc, you will also want to read up on the TF algorithm.

Dubslow · 2011-10-16, 06:05

Uhhh....
Uuuuuhhhhhhhh....
How would Java be useful? As far as I know this is all primarily C/++?

Christenson · 2011-10-16, 14:03

Java is not directly useful in the computational phase of GIMPS....but it could be used for the non-computational phases, where small numbers of compute cycles don't matter......and I was suggesting you might want to learn a bit of C before taking your Java class. That will help your development as a programmer.

2011-10-14, 02:37	#1
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 3·29·83 Posts	AVX? (Other new features?) Sorry if this is redundant somehow, but there's not been a new release since I've been around, and I'm curious how often they are. More specifically, I remember seeing in some milestone thread a while back that George mentioned he got his first AVX FFT working, so I'm wondering when that will be implemented, now that the relevant procs are out? And in this vein, when will we see mfaktc/o integration? (mfakto would make more sense.) Could it be something in the next 6 months to a year, or later or never? Thanks Dubslow

2011-10-15, 20:24	#2
cheesehead "Richard B. Woods" Aug 2002 Wisconsin USA 2²·3·641 Posts	GIMPS is an all-volunteer effort. Many of the volunteers have full-time jobs doing other things. The work will get done when it gets done. Quit asking for specificity. Are you willing to do part of the work? Then you can issue your own time estimates for your part. The source code is available for you to peruse. You could outline your qualifications to George Woltman and ask what task you might help perform. Are you willing to pay someone to get something done sooner than the all-unpaid-volunteer pace would deliver it? Last fiddled with by cheesehead on 2011-10-15 at 21:13

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Future requests?	LaurV	GPU to 72	392	2022-10-28 23:25
Upcoming features	Xyzzy	Forum Feedback	1	2007-11-26 18:57
New features in Prime95 v22.9	guido72	Software	4	2003-03-02 23:54
Desired Features/Addins	Complex33	Software	5	2002-09-05 06:13

2011-10-16, 02:04	#5
Christenson Dec 2010 Monticello 5×359 Posts	Java is going to be a tough thing to learn if it's your first programming language. Doing the communcations automagically in mfaktc is supposedly my project. It's been on the burners for six months; history suggests it will be another six months before get it finished...because of that full-time job I do, which had me working this morning in quite a bit of pain rather than sleeping in properly. The outline of the project so far has been to upgrade the worktodo parser, then I will start figuring out how to link in P95 routines from mfaktc so as to re-use as much of P95 as possible. One suggestion is you take a look at mfaktc and understand how it produces the pseudoprimes that then get passed to the GPU. We are going to want to move that task onto the GPU. Another is to help the gentleman doing the public display figure out what he wants to in terms of diagnostics out from either P95 or mfaktc. I think he's going to want to do a creative display of the FFT data being operated on. Once you have started looking at mfaktc, you will also want to read up on the TF algorithm.

2011-10-16, 06:05	#6
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 1110000110101₂ Posts	Uhhh.... Uuuuuhhhhhhhh.... How would Java be useful? As far as I know this is all primarily C/++?

2011-10-16, 14:03	#7
Christenson Dec 2010 Monticello 5·359 Posts	Java is not directly useful in the computational phase of GIMPS....but it could be used for the non-computational phases, where small numbers of compute cycles don't matter......and I was suggesting you might want to learn a bit of C before taking your Java class. That will help your development as a programmer.